Signalyst HQPlayer

I don't write reviews but whenever I find something that impresses me deeply and I think that advertising it makes world a better place. This is the reason why all my previous (the all 2 of them) are giving 5 stars. This is going to be the third one. HQPlayer is a curious tool that, like May, changed the way I see things and listen to music.

Meta

So what is it? HQPlayer comes in 3 flavors: Desktop, Embedded and Pro. Desktop is a standalone player for PC. Pro is software for producers(?) to upsample files beforehand and distribute those pre-calculated. What I'm going to focus on here is Embedded.

HQPlayer Embedded is a software that you run on your computer and it integrates with Roon seamlessly. You can connect Roon to HQPlayer from Roon's settings and then use HQPlayer as your target. Roon will send music to HQPlayer which will then upsample it and send it forward to your DAC. Note that when using HQPlayer, you should not use Roon's DSP or volume control, but deliver music to HQPlayer bit perfect and let HQPlayer do all the processing or you are going to get lower SQ.

You can configure HQPlayer from your browser (localhost:8088, if it's running on the same computer where you are trying to access it). From there you can select filters modulators, EQ, crossfeed, target DAC etc. While the web UI is somewhat ascetic (and some people have given negative comments due to that), I personally would like the dev focus on creating more filters etc.

HQPlayer has a free demo and can be used for 30min (if I recall right) at a time and then it just needs to be restarted and one can play another 30min. Anyone can try it and test it properly themselves before the purchase.

The dev

And here we get to the developer. HQPlayer is developed by a Finnish guy Jussi, who is very active on internet forums answering questions. I've also got support from him and service has been A+. He is a busy guy, but tries to answer all the questions and is very helpful. HQPlayer seems to be a passion project of his and this may explain the price of HQPlayer a bit. Compared to cost of hifi components, I would say that it's very high value purchase.

Upsampling

I would split the upsampling operation into 2 sections:

1) Filters
2) Dither (PCM) / Modulator (DSD)

There are multiple variations of both available and both the filter and dither/modulator selection affect the sound, as does the frequency to which we are upsampling. I personally would describe the effects so that the frequency defines the size of the soundstage (the higher the frequency, the bigger the space and this scales somewhat linearly the higher we go, never saturating IMO) and filter defines how it's being used. Dither/Modulator is kind of a final touch, but it's role is still substantial. On dither side, the stronger the noise shaper, the more it gives certain depth to the sound and you can hear more things happening, but at the same time I think it loses some touch to the original timbre. On modulator side, the difference is more in accuracy/revealing and certain liveliness. I think with modulators there are more clear answers on what's better, but on PCM side (dither) it's more about taste of an individual and what works with one's equipment.

In these sections I try to list the available options and how I think they affect the sound.

Filters

There are multiple different "families" of filter (gauss, ext, xlr, sinc etc) and these all have their own "house sound". Then on top of that many of these filters are available as shorter ("MQA -like") and longer ("Chord -like") versions. Shorter versions seem to emphasize macro side of things, bringing big elements into the light. Shorter versions have less what's called "ringing" and they are somewhat smoother. Longer filters on the other hand have something that are called "faster rise times" and they kind of make the show bigger and bring out small details.

Also some of the filters are "apodizing". It means that they correct errors in the poor mastering process. Jussi (the dev) has commented that most redbook material (normal CDs basically) need apodization. One can configure HQPlayer so that 44.1 and 48khz material use different filter than "real hi-res". This way one can use apodizing filter for those and something else for others.

With Utopia I clearly preferred slower filters for some reason (favourite being poly-sinc-xtr-short-lp). I think it has something to do with smaller soundstage of Utopia. With longer filters there were too many things going on and that made it hard to focus into anything. Susvara on the other hand has a big soundstage and when used with longer filters, the stage becomes epic with a lot of information. While Utopia with shorter filters converted Utopia from "in my head" into "around me in the room", Susvara with longer filters is out of this world.

Something to note: especially those longer filters tend to be quite hard to compute. I could play most combinations with my i7 8700k, but especially when I went to DSD1024, my computer started struggling. However HQPlayer supports offloading filter calculation to CUDA, ie. GPU. I've bought second hand nVidia A4000 @ 600€ and that can basically calculate anything but the most demanding monsters: sinc-Mx and sinc-L @ DSD1024. Its power draw is also quite ok when playing music.

Dither (PCM) / Modulator (DSD)

Dither is responsible for cleaning up noise (that processing produces) from music and often moving it into the inaudible frequency range. Usually the more aggressive the dither, the better it sounds (to me). All the noise dithers and noice shapers on PCM side are quite light for computer and one can just pick the one one likes most.

On modulator side there are also multiple variations available. These modulators are responsible for converting PCM into DSD. Some of the modulators that many people consider "best" (EC) tend to be heaviest to compute and even the best CPUs in the market can't compute the Holy Grail of HQPlayer: DSD1024 with ASDMECv2 modulator. Modulator calculation also cannot be outsourced to GPU and the workload can't be split for multiple cores either. That's why one needs CPU with ultra strong single core performance.

This is where a self respecting audiophiles can put their wallets into a real test (again). Most demanding modulators need the CPU with fastest single core performance there is (i9-13900KS) and most demanding filters GPU with fast 64 bit floating point calculation (things like RTX A4000 or RTX 3080 and up). For a long time the "end game" from computing perspective is sinc-L filter + ASDMECv2 modulator @ DSD1024, but the situation has changed a bit with HQPlayer 5, which introduced even more demanding filters and modulators. However what people subjectively consider best differs and if you are lucky, you may not go all the way! Since the first appearance of this review I've aquired 13900k (with AIO water cooling) and with that I can run pretty much all the modulators on DSD1024, but the two most demanding few.

Subjective testing

I start this with a warning: YMMV!

I've read some people not hearing much difference when using HQPlayer, but I've often noticed that they have conducted their tests with setups with other bottlenecks. I think HQPlayer shouldn't be an item you buy in the beginning of your audiophile career. HQPlayer is most useful when connected to a TOTL system that can reveal what it produces and it's best partnered with a NOS (non-oversampling) DAC that can rip the benefits from the highest sampling rates. You may want to consult "Recommended hardware" -section here. I've tested it with Chord Hugo 2 and the result was a bit meh. However with my current setup, it has absolutely blown my mind (especially after I bought Susvara).

The test setup

Qobuz -> Roon -> HQPlayer Embedded -> HQPlayer NAA -> Holo May -> Holo Bliss -> Focal Utopia / Hifiman Susvara
(Intel i9-13900k + nVidia A4000 to power HQPlayer, Intel NUC for NAA)

NAA is Signalyst's free software that converts any small computer into HQPlayer compatible endpoint. This way you can for example keep the potentially noisy HQPlayer computer in a garage if you want.

General Impressions

First of all HQPlayer transforms Holo May into something absolutely incredible. The size of the soundstage grows a lot and different filters give me multiple different flavors for different genres and moods, still keeping the high sound quality. Longer filters kind of stretch the soundstage and reveal all the details in the music. Shorter filters focus on macro structures in the music, keeping things better together. The effect compared to pure NOS in May is by no means small, it's huge. Even between filters and modulators the difference is very clear (especially with Susvara, less with Utopia).

Filters

Personally I prefer some filters over other. I'll list some of my favourites here and will listen to those while writing.

Filter: closed-form-M16

This is closed form interpolation (not a filter in the similar sense) with 16 million taps. I'll start with this one as I think it keeps May's own nature intact. However as it fills in the missing values, it will be easier for May to reproduce the wave form and it results in a clearer sound.

This is absolutely beautiful filter with focals as I think it has probably the best timbre and thus the "feeling" can come through. It's very nice also for active listening when you want to focus on timbre.

The downside of this filter is that it's not apodizing and thus not good fit for poorly mastered material.

Filter: poly-sinc-gauss-xla

"gauss" family is the favourite of the dev himself and this is the longest of those. It's a bit smoother than closed-form-M16, but still very "human". This is the best all-rounder IMO. Many people prefer poly-sinc-gauss-long, which is the second longest. Gauss filters work very nicely with any kind of acoustic music.

Filter: poly-sinc-ext3

"ext" family is also very special, ext2 and ext3 being the most common ones being used (ext3 again the longer one). ext3 filters are (I think) a bit smoother than gauss filters. They are less full bodied than gauss, but there is some magic to their sound. Especially for electronic music, epic soundtracks etc. It's the most "entertaining" of them all.

Filter: sinc-L

"sinc" family is closest to Chord filters (according to Jussi) and the hardest to compute. The sound is very deep with amazing separation. It's analytical and beautiful. Really nice with jazz and classical, but also with electronics with interesting acoustic scenery.

Filter: sinc-Mx

Shorter than sinc-L. Soundstage is narrower, but this one is apodizing, which makes it a better all-rounder for mixed content.

Filter: poly-sinc-xtr-short-lp

Sound becomes much more intimate, but at the same time something feels very accurate. This is by far my favourite short filter and the one I use with Focal Utopia (however with Focal Utopia I also prefer PCM upsampling). This is beautiful filter if you want music to take you away in an intimate style. Music kind of glares.

Filter: sinc-short/medium/long (new with HQPlayer 5)

This is an amazing filter. It's non-apodizing and clearly competes with sinc-L. I think this one is more accurate and requires less memory, but is much harder to compute. I can only run the short version with DSD1024, but long version with PCM (1.536MHz). I think this is technically best filter and a no-brainer for PCM users as long as it being non-apodizing is not an issue. However I'm not sure if it's always clearly better than sinc-L. It's just a different representationn

Filter: sinc-MGa (new with HQPlayer 5)

This one is really interesting filter in many ways. Firstly: it's apodizing sinc filter. Secondly: it's somewhat easy to compute. Thirdly: it's constant time, which means that the number of tap increases linearly in the function of sample rate. It's basically a "competitor" of sinc-Mx in my books, but has an edge because of the easier computation. I do think that sinc-Mx sound clearly better to my ears, but I'm not sure if that would be the case in all the systems and for all ears. Also I think sinc-MGa is a bit brighter. It may be a bit too much of a good thing with Susvara.

Modulators

Everything will be tested with closed-form-M16 filter as I think that changes less in the sound and thus it may be easier to hear differences between modulators.

Modulator: AMSDM7 512fs+

My previous go-to modulator (before obtaining 13900k) that has been designed specifically for DSD512 and higher. This is still somewhat light to process (I can still run DSD1024), but among the lighter ones produces great sound IMO. It's very accurate and lively.

Modulator: DSD7 256fs+

Also really nice modulator that is light to process. It has less focus on the treble details than AMSDM7 512fs+ and it may kick a bit better (but it's hard to verify that). If I would be forced to say something about the lower registry, I would say that DSD7 256fs+ has a bit more kick, but AMSDM7 512fs+ bass is just a tad better articulated. However I could live with either of the two and I for example like to use t his modulator with closed-form-M16 if I want to listen to a bit more relaxed presentation of it. With modulators and filters, synergies do exist.

Modulator: AMSDM7EC 512fs+

This is quickly becoming one of the two of my favorite filters. It's also a bit lighter to process than ASDM7ECv2. It's a bit like ASDM7ECv2, but somehow more open. The nature is very similar with AMSDM7 512fs+, but it has a bit more kick and character. It doesn't have the last say in timing compared to ASDM7ECv2, but it's more spacious sound and also for example human voice sounds somehow more real to me.

Modulator: ASDMECv3 (v3 new with HQPlayer 5)

The feared and loved ASDMECv3. This was updated from v2 to v3 along with HQPlayer 5 and this was a major update. The new filter is not only better sounding, but also easier to compute. My 13900k can finally handle this at DSD512 with all the filters, though I can play DSD1024x44.1 with some filters. I like to pair this with gauss-xla as that filter supports cross-family oversampling (48x -> 44.1x) and then I can conver everything to DSD1024. This has the most hifi sound of them all. I'm right now listening to trumpet and and the transients are more accurate and correct than with "sloppier" AMSDM7EC 512fs+. Music also sounds more... music. Some kind of realism is lost a bit, but it's traded to more pure sound that is still analog. ASDM7ECv3 seems to be like Chord Dave in one way. I once read in head-fi someone saying about Dave: "it's nice to do the science experiment, but after some while, you will start asking yourself the question: do I like it?". I think it depends a lot on genres, synergies and personal tastes which modulator is the best in the end. ASDM7ECv3 definitely is a good candidate and I definitely see myself playing with it a lot.

Modulator: ASDM5ECv3 (v3 new with HQPlayer 5)

This is somewhat new endeavor for me. I started listening to it last week and noticed, that I like it. I feel that it kind of leaves something in music more untouched compared to ASDM7ECv3. I would describe it so that the sound is more "naked" and maybe fragile after ASDM5ECv3. If you have experience with Chord products, I would describe it so that ASDM5ECv3 is to ASDM7ECv3 what Mojo is to Hugo 2. Sometimes less is more. It kind of lets genres like metal and rock keep their human character better and doesn't try to make everything high culture. Also with some vocals it's easier to co-live those feelings with the artist.

Modulator: ASDM7EC-super (new with HQPlayer 5)

The new light and super modulators that came with HQP 5 are very interesting and I haven't yet understood them fully. They are first of all now the hardest filters to modulators and I can only run things @ DSD512. The first thing that strikes me when I switch between super and plain ASDM7EC is space. The soundstage isn't necessarily bigger, but it creates a sense of space. I think it works through slower decay, like there was some small reverb. Also the bass is a bit sloppier with these filters compared to plain ASDM7EC.

Modulator: ASDM7EC-super 512+ fs (new with HQPlayer 5)

This is very similar to normal super filter, but it's optimized better for DSD512 sample rates and higher. Sound wise I think the biggest difference is that the holographic separation gets better, but this kind of further emphasizes the sloppiness of super filters.

There are a lot of similarities to AMSDM7EC 512+ fs as well. I think AMSDM7EC 512+ fs has similar holographic effect, but is less sloppy.

On synergies between modulators and filters

I've noticed that some modulators and filters seem to synergize better. For example ASDMECv3 + sinc-L is a combination that digs very deep and reveals everything in the sound. However, it's not always the most realistic sound and is not the best fit for human voice for example. If I want to listen to something acoustical, I tend to switch to AMSDM7EC 512fs+ with poly-sinc-gauss-xla. gauss-xla has very natural transients, losing only slightly for closed-form-M16, but it creates a better sense of space. As AMSDM7EC 512fs+ feels a bit slower than ASDMECv3, it synergizes very well with gauss-xla, while sinc-L feels sometimes kind of too fast for the AMSDM7EC 512fs+ and sound can be kind of detached (or the reverb is kind of artificial). Also I think that gauss-xla kicks harder than sinc-L and to me it's sometimes a bit too much with ASDMECv3, which is also a bit more aggressive modulator than AMSDM7EC 512fs+. In near future I will probably investigate further how gauss-xla pairs with the new super filters.

Ending notes

The domain of HQPlayer is just huge and I could just keep going on for days as the rabbit hole would probably never end (or maybe when we would start talking about overclocking CPUs with nitrogen and stuff like that). I think HQPlayer has earned this review. I've been like a child in a candy store with all the nice filters, especially since I got A4000 GPU. HQPlayer provides many beautiful variations to the system and IMO is an absolute no-brainer to try for whoever has good enough system that is able to reveal these hidden treasures.

Bravo Jussi and hurry up Intel!